EP0776136A1 - Method and apparatus for calculating a chroma key of an object evolving on a coloured background - Google Patents

Abstract

The method involves separating a subject image from a background area and defining a subject zone, a background area and a transition zone. The electronic image coordinates are transformed into polar coordinates and calculations performed (10-15). These calculations allow the size of the coloured background to be defined. The size is defined by a volume which is in the form of an open cone with an angle alpha and which has an axis of symmetry passing through a chromatic and luminance null in the colour space and a point representing the background colour.

Description

The invention relates to a method of calculating a key for cutting a subject moving in front of a colored background so as to be able to embed this subject on a new colored background.

The invention also relates to any electronic device or apparatus implementing this method such as, for example, video mixers or even autonomous devices for cutting and overlaying video images commonly known in English as "chroma-keyers". Such electronic devices or devices are used, for example, in television studio equipment.

In the following description, the source image consisting of the subject evolving on the colored background will be called subject video.

The cutting key must make it possible to better identify the subject to be cut. So it is desirable to want to keep as many details of the subject as possible, such as, for example for a character, isolated hair, the transparency of glasses, or even the smoke of a cigarette.

As is known to those skilled in the art, the cutting key has the function of allowing the distinction between the colored background of the subject video and the subject itself. It is then necessary to define in the color space a volume representing the colored background so as to be able to extract the subject from the colored background in order to embed it in front of a new background.

The invention relates to a new volume approach to the colored background so as to better integrate the details of the subject such as those mentioned above.

To this end, the invention relates to a method of calculating a key KD for cutting a subject moving in front of a colored background, said key making it possible to separate the color space into three zones, a first zone defining a volume representing the colored background, a second zone defining a volume representing the subject and a third zone defining a transition zone between the colored background and the subject. This process includes different calculation steps allowing the volume representing the colored background to be defined in the form of a cone with an opening angle α whose axis of symmetry passes through the achrome point and of zero luminance of the color space and a dot representing the color of the colored background.

Likewise, the invention relates to an electronic device comprising means for calculating a key for cutting a subject moving in front of a colored background, said key making it possible to separate the color space into three zones, a first zone defining a volume representing the colored background, a second zone defining a volume representing the subject and a third zone defining a transition zone between the colored background and the subject. The means for calculating the cutting key comprise means for calculating the coordinates of each pixel of the color space in a coordinate system (U, V, W), the coordinate system (U, V, W) being a direct trihedron whose center is the achrome point of zero luminance and whose axis W points to a point of color of the colored background, and means for defining the volume representing the colored background in the form of a cone with an opening angle α having for axis symmetry the W axis

• La figure 1 représente un exemple de volume de fond coloré selon l'art antérieur ;
• Les figures 2a et 2b représentent les transformations géométriques du référentiel (CB, CR, Y) permettant de définir le nouveau référentiel (U, V, W) de l'espace des couleurs permettant l'approche volumique du fond coloré selon l'invention ;
• La figure 3 représente un pixel de l'espace des couleurs dans le nouveau référentiel mentionné ci-dessus ;
• La figure 4 représente le volume du fond coloré selon l'invention ;
• La figure 5 représente la vue en coupe du volume du fond coloré de l'invention selon un plan perpendiculaire à un premier axe du nouveau référentiel ;
• La figure 6 représente la vue en coupe du volume du fond coloré de l'invention selon un plan perpendiculaire à un deuxième axe du nouveau référentiel ;
• La figure 7 représente l'algorithme de calcul de la clé de découpe selon l'invention ;
• La figure 8 représente un perfectionnement de l'algorithme décrit en figure 7 ;
• La figure 9 représente une description du perfectionnement décrit en figure 8.

The various characteristics and advantages of the invention will appear on reading a preferred embodiment made with reference to the appended figures among which:

• FIG. 1 represents an example of colored background volume according to the prior art;
• FIGS. 2a and 2b represent the geometric transformations of the frame of reference (CB, CR, Y) making it possible to define the new frame of reference (U, V, W) of the color space allowing the volume approach to the colored background according to the invention;
• Figure 3 shows a pixel of the color space in the new repository mentioned above;
• FIG. 4 represents the volume of the colored background according to the invention;
• FIG. 5 represents the sectional view of the volume of the colored background of the invention along a plane perpendicular to a first axis of the new reference system;
• FIG. 6 represents the cross-sectional view of the volume of the colored background of the invention along a plane perpendicular to a second axis of the new frame of reference;
• FIG. 7 represents the algorithm for calculating the cutting key according to the invention;
• FIG. 8 represents an improvement of the algorithm described in FIG. 7;
• FIG. 9 represents a description of the improvement described in FIG. 8.

FIG. 1 represents an example of colored background volume according to the prior art.

où N est le nombre de pixels du fond coloré correspondant à une fenêtre d'acquisition pointant sur tout ou partie du fond coloré.Each pixel M (n) of index n of the colored background is represented by its three coordinates yn, cbn, crn where yn, cbn and crn are respectively the luminance component, the chrominance component of red color difference and the chrominance component of difference of blue color of pixel M _(n) . For convenience, the color difference chrominance components will hereinafter be called the chrominance components. The average value y0, cb0, cr0 of each component of the colored background is then written:

where N is the number of pixels of the colored background corresponding to an acquisition window pointing to all or part of the colored background.

In the plane (CB, CR), the calculation of the Euclidean distance ρ _n of a pixel of index n with respect to the point cb0, cr0 is given by the relation: $${\text{ρ}}_{\text{not}}\text{=}\sqrt{{\left({\text{cr}}_{\text{not}}\text{- <figure-callout id="2" label="cr0" filenames="imgf0001.png,imgf0005.png" state="{{state}}">cr0</figure-callout>}\right)}^2\text{+}{\left({\text{cb}}_{\text{not}}\text{- <figure-callout id="2" label="cb0" filenames="imgf0001.png,imgf0005.png" state="{{state}}">cb0</figure-callout>}\right)}^2}\text{.}$$

Two thresholding operations, one in the plane (CB, CR) and the other along the Y axis then make it possible to define the zone of discrimination between background and subject. Thus, any point whose luminance value is between the extreme values y _min and y _max and whose Euclidean distance is less than a value ρ ₁ belongs to the colored background 1.

Similarly, any point whose luminance value is between the values y _min and y _max and whose Euclidean distance ρ _n is greater than a value ρ ₂ , itself greater than ρ ₁ , belongs to subject 3. Finally, any point whose luminance value is between the values y _min and y _max and whose Euclidean distance ρ _n is between ρ ₁ and ρ ₂ belongs to an intermediate transition zone 2.

Adjusting the content of the image proves difficult with such a volume approach. The cylindrical shape of the volume described in Figure 1 has many drawbacks. For example, it is difficult to deal with too large variations in luminance of the colored background without affecting the luminance of the subject. Similarly, the pixels representing the drop shadow cannot be integrated into the volume of the colored background. It follows that any treatment of the shade proves to be impossible.

The invention does not have these drawbacks.

Figures 2a and 2b show the different transformations used to build the volume defining the colored background according to the invention.

où cb et cr sont les composantes d'un point de l'espace des couleurs selon les axes CB et CR où u et z sont les composantes de ce même point selon les axes U et Z.FIG. 2a represents a rotation of the plane (CB, CR) at zero luminance (Y = 0) around the axis Y. The angle of rotation β is such that the axis CR becomes the axis Z which points in the direction D towards the hue of the color of the colored background. Using this same rotation, the axis CB becomes the axis U perpendicular to the axis Z so that the trihedron (U, Z, Y) is a direct trihedron. This change of reference can be expressed in matrix form by the relation:

where cb and cr are the components of a point in the color space along the axes CB and CR where u and z are the components of this same point along the axes U and Z.

Figure 2b represents a rotation of the plane (Z, Y) around the axis U. The angle of rotation γ is such that the axis Y becomes the axis W which points in the direction C towards the color of the colored background . Using this same rotation, the Z axis becomes the V axis perpendicular to the W axis so that the trihedron (U, V, W) is a direct trihedron.

où z et y sont les composantes d'un point de l'espace des couleurs selon les axes Z et Y et où w et v sont les composantes de ce même point selon les axes W et V.This change of reference can be expressed in matrix form by the relation:

where z and y are the components of a point in the color space along the Z and Y axes and where w and v are the components of this same point along the W and V axes

FIG. 3 represents a pixel of the color space in the new reference frame constructed in FIGS. 2a and 2b.

The pixel M _(i) has a component along each of the three axes U, V, W, that is, respectively, the components u _(i) , v _(i) and w _(i) .

et $${\text{θ}}_{\text{(i)}}\text{= (}\overrightarrow{\text{U}}\text{,}\overrightarrow{\text{O}}{\text{P}}_{\text{(Mi)}}\text{)}$$

où le point P_(Mi) est la projection du point M_(i) sur le plan (U, V).The cylindrical coordinates of the point M _(i) in the coordinate system (U, V, W) are: r _(i) , θ _(i) and w _(i) . As is known, it comes: $$\text{r}{}_{\left(\text{i}\right)}\text{=}\sqrt{\text{<figure-callout id="2" label="u i" filenames="imgf0001.png,imgf0005.png" state="{{state}}">u </figure-callout>}{}_{\left(\text{i}\right)}{\left(\right)}_{}{}^{\text{2}}\text{+ <figure-callout id="2" label="v i" filenames="imgf0001.png,imgf0005.png" state="{{state}}">v </figure-callout>}{}_{\left(\text{i}\right)}{\left(\right)}_{}{}^{\text{2}}}$$

and $${\text{θ}}_{\text{(i)}}\text{= (}\overrightarrow{\text{U}}\text{,}\overrightarrow{\text{O}}{\text{P}}_{\text{(Mid)}}\text{)}$$

where point P _(Mi) is the projection of point M _(i) on the plane (U, V).

In the following description, the distance r _(i) will be called the chrominance distance of the point M _(i) relative to the axis W.

FIG. 4 represents the volume approach of the colored background according to the invention.

The volume which represents the colored background has the shape of a whole cone (volume V1) or truncated (volume V2) in the color space referenced with respect to the new frame (U, V, W) deduced from the frame (CB, CR, Y) by the geometric transformations described in Figures 2a and 2b. The color space is divided into three zones, as before: zone 1 which represents the colored background, zone 3 which represents the subject and zone 2 which represents the transition between the subject and the colored background.

The cone representing the colored background is defined by an angular opening α, a vertex located at the abscissa ws on the axis W, an elliptical or circular section. A high luminance threshold, of abscissa wh on the W axis, makes it possible to define the face opposite to the top of the cone. A low luminance threshold, of abscissa wb on the axis W, makes it possible to define the truncated face situated at the distance d from the vertex of abscissa ws. The abscissa wb can coincide with the abscissa ws. According to the invention, the cone can rotate by an angle R varying from 0 to π around the axis W.

The transition zone 2 is defined by the space between the surface of the opening cone α and the surface of an opening cone α 1 greater than α, with the same axis of symmetry and the same vertex as the cone d opening α and whose face opposite the top of the cone is in the same plane as the face opposite the top of the opening cone α.

The area 3 representing the subject is defined by the space located beyond the opening cone α 1.

FIG. 5 represents the sectional view of the volume of the colored background of the invention according to a plane (U ₀ , V ₀ ) of cross section of the cone.

This sectional view shows an elliptical section. This elliptical section is capable of rotating around the axis W by an angle R as mentioned above. The coordinates r _(i) and θ _(i) of each pixel M _(i) delimiting the colored background are then connected by the equation: $${\text{k}}^{\text{2}}{\text{r}}_{\text{(i)}}{}^{\text{2}}{\text{cos}}^{\text{2}}{\text{(θ}}_{\text{(i)}}{\text{+ R) + r}}_{\text{(i)}}{}^{\text{2}}{\text{sin}}^{\text{2}}{\text{(θ}}_{\text{(i)}}\text{+ R) = constant.}$$

The modulation of the parameter k allows the deformation of the ellipse. According to the invention, the elliptical modulation of the distance chrominance leaves the possibility of making a circle in the case where k = 1. Preferably k varies between 1 and 2. The value of R can vary from 0 to π. Advantageously, this modulation technique makes it possible to increase the selectivity of the volume defining the colored background. For example, the elliptical modulation technique associated with the rotation R allows a clear improvement to reproduce the transparency of the glasses.

Figure 6 shows the sectional view of the volume of the colored background of the invention along a plane passing through the major axis of the ellipse and the top of the cone.

où CLIP est un paramètre dont la variation permet d'engendrer le déplacement de l'ordonnée ws du sommet du cône sur l'axe W. A titre d'exemple, la plage de variation de l'angle α est comprise entre 0 et 45°. L'action conjuguée des paramètres α et CLIP permet avantageusement une modulation de la définition du fond coloré, et partant du sujet à traiter. Pour une valeur de seuillage ws nulle, l'ensemble du volume découpé descend jusqu'au plan des noirs.According to the invention, the conical volume is defined so that the elliptically modulated chrominance distance, denoted r _e (i), the aperture angle α of the cone and the coordinate w (i) of the point whose chrominance distance is r _e (i) are connected by the equation: $${\text{r}}_{\text{e}}\text{(i) - α w (i) = CLIP}$$

where CLIP is a parameter whose variation makes it possible to generate the displacement of the ordinate ws from the apex of the cone on the axis W. For example, the range of variation of the angle α is between 0 and 45 °. The combined action of the parameters α and CLIP advantageously allows modulation of the definition of the colored background, and therefore the subject to be treated. For a threshold value ws zero, the entire cut volume goes down to the black plane.

The variation of the abscissa ws makes it possible to integrate into the volume which defines the colored background points of more or less weak luminance having the tint of the colored background. Advantageously, it is then possible to take into account non-uniform colored backgrounds, the luminance of which varies considerably to the eye. As mentioned previously (see Figure 4), the cone can be truncated by a distance d in the case where wb is different from ws.

The CLIP parameter is used to calculate the chrominance key KC. Any point in the image whose chrominance distance is less than the CLIP value belongs to the colored background 1 and has a zero chrominance key KC.

The transition zone 2 is associated with the thresholding of the chrominance distance and makes it possible to effect the gradual transition between the background and the subject. The size of the transition zone is adjustable by an adjustment parameter called GAIN. To this end, the GAIN parameter makes it possible to calculate the distance separating the points belonging to the aperture cone α1 from those belonging to the aperture cone α. Any point in the image located in zone 3, beyond the transition zone 2, belongs to the subject and has a chrominance key value KC equal to 1.

All the abscissa points w (i) lower than wb have a luminance key KL equal to 1. All the abscissa points w (i) greater than wh have a luminance key KL equal to 0. All the points d 'abscissa w (i) located between wb and wh are in the luminance transition zone.

The generation of the key KD for the colored background is carried out thanks to the combination of the chrominance key KC and the luminance key KL. Thus, for each of the points of the image, the calculation of the maximum of the two keys allows the construction of the volume defining the colored background; he comes : $$\text{KD = MAX [KC, KL].}$$

FIG. 7 represents the algorithm for calculating the cutting key according to the invention. This algorithm synthesizes the different stages preferably performed by microprocessor and described in FIGS. 2a and 2b in FIG. 6.

Each pixel of the source video VS comprises three components cbs, crs, ys in the color space referenced by the reference (CB, CR, Y).

As is known to those skilled in the art, the luminance and chrominance data are generally in 4: 2: 2 format. The sampling frequency of the luminance components is 13.5 MHz while that of the components of chrominance is 6.75 MHz. It is then necessary to interpolate the train of chrominance components cbs, crs which is in 2: 2 format into a train of chrominance components cb1, cr1 in 4: 4 format. The interpolation operator 10 provides for this. Preferably, the chrominance components of the pixels of the video source VS are interrupted during line deletions so that the interpolated components located at the start and end of the line are not disturbed by the line delete signal. The interpolation is carried out in a manner known per se. The information relating to the chrominance components cbs and crs is of integer type. In order to simplify the interpolation calculation, the achrome level which corresponds, for example, to level 512 for a video signal coded on 10 bits is converted into arithmetic zero by inversion of the value of the most significant bit. By way of example, the interpolation can be carried out on 55 color difference points so as to obtain very high calculation precision. At the output of the interpolator 10, the components cb and cr are coded on a number N of bits greater than the number of bits of the video. Thus, for a video signal coded on 10 bits at the input of the interpolator, can the dynamics of the samples cb, cr obtained at the output of the interpolator, for example, be coded on 16 bits among which are the 10 bits of the video, 4 bits of fractionation intended to define the precision of the video, 1 bit of sign and 1 bit of overflow.

The chrominance components cb and cr resulting from the interpolation calculation are transformed into chrominance components u and z by horizontal rotation 11 of angle β such as that described in FIG. 2a.

The horizontal rotation is carried out in two stages. First, the microprocessor loads the sin (β) and cos (β) information during frame deletion and a multiplier performs the calculation of the products cr x cos (β), cr x sin (β), cb x cos ( β) and cb x sin (β), for each point of the image, during the active frame. The u and z information travels at a frequency of 13.5 MHz. The information cos (β), sin (β) circulates at the frequency of 27 MHz. It follows that at the output of the multiplier the information is multiplexed at the frequency of 27 MHz.

Secondly, the information multiplexed at the frequency of 27 MHz is demultiplexed and matrixed to obtain the signals u and z at the frequency of 13.5 MHz.

The matrix relation, already mentioned during the description of FIG. 2a, is written:

The luminance samples ys of the subject video Vs are delayed by the delay 19 in luminance samples y so as to compensate for the delay taken by the chrominance samples cbs and crs during the interpolation 10 and horizontal rotation operations 11. In order for the luminance information to have a format identical to the chrominance information cb, cr, the luminance samples are translated from scales so as to match the luminance zero with the arithmetic zero and are coded on a number N of bits identical to that of the chrominance samples. The components cb, cr, y of each pixel are then transformed into components (u, v, w) by vertical rotation of angle γ such as that described in FIG. 2b.

The vertical rotation is carried out according to the same principle as that described for the horizontal rotation.

First, the microprocessor loads the sin (γ) and cos (γ) information during the frame suppression while a multiplier performs the calculation of the products zx cos (γ), zx sin (γ), yx cos (γ ) and yx sin (γ) during the active frame. Information z, circulates there at a frequency of 13.5 MHz. The information cos (γ), sin (γ) circulates at the frequency of 27 MHz. It follows that at the output of the multiplier, the information is multiplexed at the frequency of 27 MHz.

In a second step, the information multiplexed at the frequency of 27 MHz is demultiplexed and matrixed to obtain the signals w and v at the frequency of 13.5 MHz. The matrix relation, already mentioned during the description of FIG. 2b, is written:

During the vertical rotation operation 12, the signal w is delayed so as to compensate for the processing time necessary for the operation 13 for calculating the coordinates r, θ which follows the vertical rotation operation 12.

$$\text{θ =}\overrightarrow{\text{U}}\text{,}\overrightarrow{\text{O}}{\text{P}}_{\text{(M)}}\text{),}$$

comme cela a été mentionné précédemment.The coordinates r and θ associated with each pixel are given by the formulas: $$\text{r =}\sqrt{{\text{<figure-callout id="2" label="u" filenames="imgf0001.png,imgf0005.png" state="{{state}}">u</figure-callout>}}^{\text{2}}{\text{+ <figure-callout id="2" label="v" filenames="imgf0001.png,imgf0005.png" state="{{state}}">v</figure-callout>}}^{\text{2}}}$$

$$\text{θ =}\overrightarrow{\text{U}}\text{,}\overrightarrow{\text{O}}{\text{P}}_{\text{(M)}}\text{),}$$

as previously mentioned.

The coordinates r and θ are recovered at the output of the operator 13 in order to carry out the elliptical modulation 14.

The elliptical modulation 14 is preferably carried out by the calculation of a modulating factor A. The factor A is calculated for each pixel, during the frame blanking period, using two adjustment parameters k and R such as those mentioned during the description of Figure 5; he comes : $$\text{A =}\sqrt{{\text{<figure-callout id="2" label="k" filenames="imgf0001.png,imgf0005.png" state="{{state}}">k</figure-callout>}}^{\text{2}}{\text{cos}}^{\text{2}}{\text{(θ + R) + sin}}^{\text{2}}\text{(θ + R).}}$$

The factor k allows the elliptical deformation of the chrominance distance r and the angle R allows the rotation of the elliptical section of the cone from 0 to π.

The elliptically modulated chrominance distance is then written $${\text{r}}_{\text{e}}{\text{= r (A)}}^{\text{1/2}}\text{.}$$

The calculation of A is carried out, for example, using EPROMs wired for this purpose. The dynamics of the multiplication of r by (A) ^1/2 takes place on N bits, for example 16 bits, at the sampling frequency of the luminance components, ie 13.5 MHz.

The elliptical modulation 14 is followed by an operation 15 allowing the opening of the cone and the definition of the cutting volume using the parameters α, CLIP, GAIN, wb and wh.

The conical aperture angle α as well as the chrominance distance threshold CLIP value are applied to the values r _e and w of each pixel according to the law previously stated in FIG. 6, that is: $${\text{r}}_{\text{e}}\text{- α w = CLIP.}$$

The GAIN parameter is used to define the size of the transition zone between the colored background and the subject as mentioned above.

Likewise, the parameters wb and wh, respectively low luminance threshold and high luminance threshold, make it possible to delimit the volume of the colored background along the axis W.

Thus, an advantage of the invention is the easily adaptive nature of the volume defining the colored background.

The definition of the cutting volume allows the generation of the chrominance key KC and the luminance key KL as mentioned above.

The keys KC and KL resulting from the operation 15 for defining the cutting volume are then combined by a combining operation 16 so as to obtain the key KD allowing the cutting of the volume defining the colored background. As mentioned earlier, it comes: $$\text{KD = MAX [KC, KL].}$$

The pixels for which KD = 1 belong to the subject and those for which KD = 0 belong to the colored background. For the pixels of the transition zone, KD is between 0 and 1.

According to the preferred embodiment of the invention, the operation 16 of combining the keys KC and KL is followed by an operation 17, known per se, of masking and filtering the key KD. Indeed, during the extraction of the subject, it happens that a part of the details which belong to the subject is assimilated to the colored background. The most frequently encountered example is the loss of colorimetry of a character's eyes. The introduction of the masking and filtering operation then locally invalidates the level of the cut. For example, the mask can be rectangular in shape and adjustable in size. However, it is possible to use different mask shapes from shape generators.

As is known to those skilled in the art, the filtering of the cutting key makes it possible to eliminate the color residues from the background of the subject video.

The subject's final cutting key KDF comes from the masking and filtering operation 17. The KDF key is then applied in a known manner to the final mixing operation to generate the image of the subject moving in front of the new background.

FIG. 8 represents an improvement of the algorithm described in FIG. 7.

This improvement concerns the generation of DZ zone detection information from an operator 19.

Advantageously, this improvement allows the detection of the shadow cast by the subject on the colored background. The drop shadow is characterized by luminance information lower than the average luminance level of the colored background and of the same shade as that of the colored background.

According to the invention, the shadow cast can be removed or restored by dimming the luminance of the new reception background. For the information that allows cutting, the shadow is not part of the colored background. However, it must be distinguished from the subject. To this end, any pixel located around the tint of the colored background, whose luminance value is less than a certain threshold and which belongs to the transition zone 2 situated between the colored background 1 and the subject 3 is considered to be a shadow pixel.

Advantageously, the orientation and the conical shape of the volume defining the colored background allow the shadow pixels to enter the transition zone 2 separating the subject from the colored background.

In FIG. 8 are shown the interpolation operator 10, the horizontal rotation operator 11, the delay operator 19 and the masking and filtering operator 17, as described in FIG. 7. For reasons of simplification, the operators 12, 13, 14, 15 and 16 are grouped in the same block B.

According to the refinement of the invention, the operator 19 for detecting areas includes operations allowing the detection of the shadow cast by the subject on the colored background. The luminance component y originating from the delay operator 18 is compared with a threshold value of luminance ya and the angular coordinate θ from the operator 13 described in Figure 7 is compared to two values θmin and θmax whose difference θmax - θmin defines the angular opening in which the angle θ must be located so that the hue of the pixel associated with θ is assimilated to the tint of the colored background. Any pixel whose value of θ is between θ min and θmax and whose luminance value y is less than ya is considered as a shadow pixel.

As mentioned previously, the shadow pixels are located in the transition zone 2 separating the subject from the colored background. Thus, the key KD is applied to the operator 19 of zone detection so as to provide the information making it possible to locate the pixels of the transition zone 2. The signal DZ coming from the operator 19 of zone detection then makes it possible to give information on the detection of shadow pixels according to the method mentioned above.

Advantageously, as it appears in FIG. 9, the DZ signal can provide not only information for locating the shadow pixels but also various information making it possible to locate the pixels of the subject video according to their position in the color space.

In FIG. 9 are shown the different stages of the operation 19 of zone detection in the form of a flow diagram in the case where the signal DZ provides all. the information making it possible to locate the pixels of the subject video according to their position in the color space. The invention however also relates to the cases where DZ gives all or part of this information.

If the key KD has the value 1, the information DZ takes a value x1 reflecting the detection of the subject.

If the key KD has the value 0, the information DZ takes a value x2 translating the detection of the background.

• Si y n'est pas inférieur à ya, l'information DZ prend une valeur x3 traduisant la détection de la zone de transition entre fond et sujet.
• Si y est inférieur à ya et si θ est compris entre θmin et θmax, l'information DZ prend une valeur traduisant la détection de l'ombre. Préférentiellement, cette valeur est égale à x2 de façon que les pixels d'ombre soient traités comme les pixels du fond.
• Si y est inférieur à ya et si θ n'est pas compris dans l'intervalle θmin, θmax, l'information DZ prend une valeur traduisant la détection de la zone de transition entre fond et sujet.

If the key KD has a value between 0 and 1, the luminance value y is compared with the threshold value ya and:

• If y is not less than ya, the DZ information takes a value x3 reflecting the detection of the transition zone between background and subject.
• If y is less than ya and if θ is between θmin and θmax, the DZ information takes a value reflecting the detection of the shadow. Preferably, this value is equal to x2 so that the shadow pixels are treated like the background pixels.
• If y is less than ya and if θ is not included in the interval θmin, θmax, the information DZ takes a value translating the detection of the transition zone between background and subject.

Claims (25)

Method for calculating a KD key for cutting a subject moving in front of a colored background, said key making it possible to separate the color space into three zones, a first zone (1) defining a volume representing the colored background, a second zone (3) defining a volume representing the subject and a third zone (2) defining a transition zone between the colored background and the subject, characterized in that it comprises different calculation steps (10, 11, 12, 13, 14, 15) used to define the volume representing the colored background in the form of an aperture cone α whose axis of symmetry (W) passes through the achrome point and of zero luminance of the color space and a dot representing the color of the colored background. Method according to claim 1, characterized in that the polar coordinates r _(i) and θ _(i) of the different pixels (Mi) which define a cross section of the cone in a plane (Uo, Vo) such as the trihedron (Uo, Vo, W) or a direct trihedron, follow the equation: $${\text{k}}^{\text{2}}{\text{r}}_{\text{(i)}}{}^{\text{2}}{\text{cos}}^{\text{2}}{\text{(θ}}_{\text{(i)}}{\text{+ R) + r}}_{\text{(i)}}{}^{\text{2}}{\text{sin}}^{\text{2}}{\text{(θ}}_{\text{(i)}}\text{+ R) = constant,}$$

where θ _(i) is the polar angle defining the point M _(i) in the plane (Uo, Vo), k and R being respectively an elliptical modulation factor of the coordinate r _(i) and an angle of rotation of the cross section of the cone. Method according to claim 2, characterized in that the aperture α of the cone is connected to a CLIP adjustment parameter so that: $$\text{r}{}_{{}_{\text{e}}\text{(i)}}{\text{- α w}}_{\text{(i)}}\text{= CLIP}$$

or $$\text{r}{}_{{}_{\text{e}}\text{(i)}}{\text{= r}}_{\text{(i)}}{\text{(AT)}}^{\text{1/2}}$$

with $$\text{A =}\sqrt{{\text{k}}^{\text{2}}{\text{cos}}^{\text{2}}{\text{(θ}}_{\text{(i)}}{\text{+ R) + sin}}^{\text{2}}{\text{(θ}}_{\text{(i)}}\text{+ R)}}$$

and where w _(i) is the coordinate on the axis of symmetry (W) of said cone of the plane (Uo, Vo), the face opposite the apex of the cone being defined by a coordinate wh on the axis of symmetry (W) of the cone. Method according to claim 2 or 3, characterized in that the elliptical modulation factor k is such that 1 ≤ k ≤ 2, and in that the angle R can vary from 0 to π. Method according to any one of Claims 1 to 4, characterized in that the transition zone between the colored background and the subject is defined by an area of space comprised between the surface of the opening cone α defining the colored background and an aperture cone α1 greater than α, with the same axis of symmetry and the same vertex as the cone defining the colored background, and whose face opposite the apex is in the same plane as the face opposite the apex of the cone defining the bottom colored. Method according to claim 5, characterized in that the size of the transition zone is adjustable by an adjustment parameter (GAIN). Method according to any one of Claims 1 to 6, characterized in that the cone defining the colored background can be truncated by a distance d separating the coordinate ws giving the position of the apex of the cone on the axis of symmetry W d ' a coordinate wb defining, on said axis of symmetry, a low luminance threshold wb having a value between ws and wh. Method according to one of claims 2 to 6, characterized in that the cutting key KD comes from an operation of combining a chrominance key KC and a luminance key KL so that: $$\text{KD = MAX [KC, KL]}$$

the luminance key KL having the value 1 for any pixel whose coordinate w (i) on the axis of symmetry of the opening cone α is less than wb and for value 0 for any pixel whose coordinate w (i) on the axis of symmetry α of the cone of aperture is greater than wh, and the chrominance key KC having the value 0 for any pixel belonging to the cone aperture α and the value 1 for any pixel situated beyond the cone 'opening α 1. Method according to any one of the preceding claims, characterized in that the achrome level of the various chrominance components and the zero level of the luminance component which constitute the origin of the color space marker are converted into arithmetic zeros. Method according to any one of the preceding claims, characterized in that it comprises an interpolation step (10) so that the chrominance components of red color difference (crs) and blue color difference (cbs) of the pixels of the image (VS) formed by the subject moving in front of the colored background are transposed to the same video format as the luminance components of said pixels. Method according to claim 10, characterized in that it comprises a step making it possible to code the chrominance component of red color difference (crs), the chrominance component of blue color difference (cbs), and the luminance component (ys) of each pixel on a number of bits N greater than the number of video bits of said components. Method according to claim 11, characterized in that the number of video bits of the red and blue color difference chrominance components and of the luminance components is equal to 10 and in that the number N of bits on which said chrominance components are coded and luminance is equal to 16, among which are the 10 bits of the video, 4 splitting bits intended to define the precision of the video, 1 sign bit and 1 overflow bit. Method according to any one of the preceding claims, characterized in that it comprises a step of detecting zones making it possible to deliver DZ information giving the location of the different pixels of the image constituted by the subject moving in front of the colored background. Method according to claim 13, characterized in that the zone detection step comprises a step making it possible to deliver DZ information giving the location of the pixels representing the shadow of the subject cast on the colored background. Method according to claim 14, characterized in that the step making it possible to deliver the DZ information giving the location of the pixels representing the subject's shadow cast on the colored background comprises the following sequence: - If the cutting key KD is between 0 and 1, the luminance value y of each pixel M (i) is compared with a threshold value ya and: - If y is greater than ya, the DZ information takes on a value reflecting the detection of the transition zone between background and subject. - If y is less than ya and if the angle θ _(i) is between two values θmin and θmax whose difference θmin - θmax defines the angular opening in which the angle θ _(i) must be located so that the tint of the pixel associated with θ _(i) is assimilated to the tint of the colored background, the DZ information takes on a value reflecting the detection of the shadow. - if y is less than ya and if the angle θ _(i) is not between said values θmin and θmax, the information DZ takes a value reflecting the detection of the transition zone between background and subject. Method according to any one of the preceding claims, characterized in that it comprises an operation (17) for masking and filtering making it possible to calculate a final cutting key (KDF) from the cutting key KD. Electronic device comprising means for calculating a key (KD) for cutting a subject moving in front of a colored background, said key (KD) making it possible to separate the color space into three zones, a first zone (1) defining a volume representing the colored background, a second zone (3) defining a volume representing the subject and a third zone (2) defining a transition zone between the colored background and the subject, characterized in that the means for calculating the key (KD) cutting means comprise means (10, 11, 12) for calculating the coordinates of each pixel of the color space in a coordinate system (U, V, W), the coordinate system (U, V, W) being a direct trihedron of which the center is the achrome point of zero luminance and the W axis of which points to a color point of the colored background, and means (13, 14, 15) for defining the volume representing the colored background in the form of a corner cone aperture α having as axis of symmetry the axis W. Apparatus according to claim 17, characterized in that it comprises means (14) for elliptically modulating the cross section of the cone with an opening angle α . Apparatus according to any one of claims 17 and 18, characterized in that it comprises means (GAIN) so that the transition zone between the colored background and the subject is defined by an area of space between the surface of the aperture cone α defining the colored background and an aperture cone α 1 greater than α, with the same axis of symmetry and the same vertex as the cone defining the colored background, and whose face opposite the apex is in the same plane as the face opposite the top of the cone defining the colored background. Apparatus according to any one of claims 17 to 19, characterized in that it comprises means for truncating by a distance d the cone defining the colored background, said distance d separating the coordinate ws giving the position of the top of the cone on the axis of symmetry W of a coordinate wb defining, on said axis of symmetry, a low luminance threshold wb having a value between ws and wh, wh being the abscissa of the plane defining the face opposite the apex of the cone. Apparatus according to any one of claims 17 to 20, characterized in that it comprises means making it possible to detect said zones (1, 2, 3) in order to deliver information (DZ) giving the location of the different pixels of the image formed by the subject moving in front of the colored background. Apparatus according to claim 21, characterized in that the means making it possible to detect said areas (1, 2, 3) comprise means for delivering information (DZ) giving the location of the pixels representing the shadow of the subject cast on the colored background . Apparatus according to any one of claims 17 to 22, characterized in that it comprises means for embedding the subject cut out on a new reception background. Video mixer, characterized in that it comprises an apparatus according to any one of claims 17 to 23. Autonomous device for cutting and overlaying a video image, characterized in that it comprises an apparatus according to any one of claims 17 to 23.

Images